Feature Extraction Methods On A Partial Section Of The Iris Region For Iris Classification

Iris classification is a biometric system to classify a person using the individual’s iris pattern. One of the important steps in this system is to extract the iris information from the segmented iris region. Although several methods have produced a perfect recognition rate, they require intensive p...

Full description

Saved in:
Bibliographic Details
Main Author: Ali, Ahmad Nazri
Format: Thesis
Language:English
Published: 2019
Subjects:
Online Access:http://eprints.usm.my/56376/1/Feature%20Extraction%20Methods%20On%20A%20Partial%20Section%20Of%20The%20Iris%20Region%20For%20Iris%20Classification_Ahmad%20Nazri%20Ali.pdf
http://eprints.usm.my/56376/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Iris classification is a biometric system to classify a person using the individual’s iris pattern. One of the important steps in this system is to extract the iris information from the segmented iris region. Although several methods have produced a perfect recognition rate, they require intensive processing that involves the process of isolating the iris information as well as other information such as eyelid and eyelashes during template generation. The process of separating these two parts is crucially needed so that no eyelid or eyelash data are acknowledged as iris data during matching. To define the issue, the widely used approach of the feature extraction method as proposed by Daugman is studied in this research work. Then, an alternative feature extraction technique by using the upper half of the iris region that is able to skip the process of separating between iris information and eyelids or eyelashes during feature computation is proposed which is not only able to reduce the computation time but is able to preserve the accuracy rate. The proposed schemes are based on difference cumulative bin (DCB), sequential cumulative bin (SCB) and overlap mean intensity (OMI) that utilize the local texture analysis computation for transforming pixel value to a binary bit. The methods are assessed using Support Vector Machines (SVM), k-NN and Naïve Bayes classifiers on various region sizes and neighbourhood elements. The result showed that although the average accuracy for the proposed methods on individual assessment (94.27%) was slightly lower than by the Daugman method (95.77%), the classification rate for the proposed methods has improved to achieve 96.26% accuracy if the assessment uses a concatenated mode set of features and also has managed to reduce the computation time which is 0.030 ms compared to Daugman’s method that required 0.166 ms.